Presentation Information
[18a-PA1-16]Inverse Design of Solid Polymer Electrolytes Using Causal Discovery and Generative AI
〇Meguru Yamazaki1, Yuta Yoshimoto1, Eiji Ohta1, Naoki Matsumura1, Kazutaka Nishiguchi1, Hiroyuki Higuchi1, Yasufumi Sakai1 (1.Fujitsu Limited)
Keywords:
Structure generation,Inverse design,Causal Discovery
Several models have been proposed to generate structures with desired features. However, identifying which features effectively enhance those properties still relies heavily on expert knowledge. In this study, we introduce a method that integrates generative model with causal discovery to automatically extract relevant features from literature. We evaluated the approach on solid polymer electrolytes and successfully identified functional groups contributing to improved lithium-ion conductivity. Structures containing these groups were generated and validated through MD simulations, confirming that they achieve the targeted property.